Article Preview
TopIntroduction
In the summer of 2005, after a routine review of case practice that involved a set of child fatalities, the New York State Commissioner for the Administration for Children Services (ACS) identified several areas in child protection operations, policy, and practice that needed to be changed or strengthened to improve service delivery. Another high-profile child fatality that occurred in January 2006 created additional political and public pressure that resulted in ACS releasing an improvement plan that included a series of initiatives to strengthen case practice, reassess and improve policy and procedures, and infuse resources directly to the front line. It also called ACS to develop ChildStat, an information management system within the family of PerformanceStat systems (Behn, 2003). PerformanceStat is a good characterization of performance strategies in the public sector that use relatively real-time data, information and communication technologies, performance measures, and mechanisms to bring together different actors to make decisions, manage, or learn about a government program (DeHaven-Smith & Jenne II, 2006). Using information and communication technologies to manage and process performance information is becoming a common strategy in performance management (Rochet et al., 2009; Schooley & Horan, 2007; Yang & Maxwell, 2011; Yetano et al., 2021).
A key component of many PerformanceStat systems consists of the use of dialogue to promote learning in performance management (Bebbington et al., 2007; de Bruijn & van Helden, 2006; Laihonen & Mäntylä, 2017; Lewis & Triantafillou, 2012; Moynihan, 2005). Unfortunately, as observed in the ChildStat case, the impact of dialogue is not deterministic or easy to predict, leading sometimes to coordinated action, and sometimes to crystallize conflict. Research in performance dialogue has found that dialogue can lead to pseudo-participation (Aleksandrov et al., 2018), and that problems may emerge when performance dialogue is not linked to decision making (Laihonen & Mäntylä, 2017). Researchers like Lewis and Triantafillou (2012) even suggest to be cautious given that performance dialogue and other learning approaches to performance management may lead to government overload. In addition, although current performance dialogue systems have been implemented with some success, blocked communication has been identified as a main problem of performance dialogue (DeHaven-Smith & Jenne II, 2006). In this way, this paper is guided by the following research question: what are main learning processes and information involved in performance dialogue? And how this information and learning processes may explain communication difficulties in performance dialogue?
Both questions have been identified as important research directions in performance management given that answers will contribute to (1) understanding the effects of different information types and representations on performance information use (Brown & Dillard, 2015b; Kroll, 2015), and (2) understanding how to better design dialogue systems that lead to program improvement and learning (Aleksandrov et al., 2018; de Bruijn & van Helden, 2006). The paper also contributes to a better understanding on how to use information systems for performance management beyond implementation, which has been identified as an important research direction (Brown & Dillard, 2015b; Yetano et al., 2021).
To understand performance dialogue, an interdisciplinary approach is necessary (Pandey, 2016). A great deal of performance studies have found shortcomings related to performance information and performance information use, which is plagued by multiple interpretations and meanings (Kroll, 2015, 2016; Talbot, 2005). To contribute to this literature, we approach the problem of performance information use as knowledge sharing across boundaries. Moreover, we suggest that performance information can be understood as boundary objects (Black et al., 2004; Carlile, 2002; Star & Griesemer, 1989), which are defined as physical and non-physical configurations of metrics, data, visualizations, and other performance-related information that are used in performance management. Our analysis suggest that interactive dialogue can have a failing or success mode as a result of blocked communication, which can be explained –at least in part—by the types and formats of performance information as a boundary object. We propose a conceptual model of knowledge boundaries that capture knowledge sharing using performance information in different contexts, and show that performance information characteristics are key determinants of the success and failure modes of performance dialogues.